Dedicated to a memory of remarkable mathematician and man Victor Petrovich Havin ON THE MAXIMUM PRINCIPLE FOR THE RIESZ TRANSFORM VLADIMIR EIDERMAN AND FEDOR NAZAROV Abstract. Let µ be a measure in Rd with compact support and continuous density, and let y−x 7 Rsµ(x)= dµ(y), x,y ∈Rd, 0<s<d. Z |y−x|s+1 1 0 We consider the following conjecture: 2 sup |Rsµ(x)|≤C sup |Rsµ(x)|, C =C(d,s). n x∈Rd x∈suppµ a This relationwas knownfor d−1≤s<d, and is still an open problem in the generalcase. J We prove the maximum principle for 0<s<1, and also for 0<s<d in the case of radial 7 measure. Moreover,we show that this conjecture is incorrect for non-positive measures. 1 ] A C 1. Introduction . h Let µ be a non-negative finite Borel measure with compact support in Rd, and let 0 < s < t a d. The truncated Riesz operator Rs is defined by the equality m µ,ε [ y −x Rs f(x) = f(y)dµ(y), x,y ∈ Rd, f ∈ L2(µ), ε > 0. 1 µ,ε Z |y −x|s+1 |y−x|>ε v 0 For every ε > 0 the operator Rs is bounded on L2(µ). By Rs we denote a linear operator µ,ε µ 0 on L2(µ) such that 5 y −x 4 Rsf(x) = f(y)dµ(y), 0 µ Z |y −x|s+1 . 1 whenever the integral exists in the sense of the principal value. We say that Rs is bounded 0 µ 7 on L2(µ) if 1 kRsk := supkRs k < ∞. : µ µ,ε L2(µ)→L2(µ) v ε>0 Xi In the case f ≡ 1 the function Rs1(x) is said to be the s-Riesz transform (potential) of µ µ r and is denoted by Rsµ(x). If µ has continuous density with respect to the Lebesgue measure a m in Rd, that is if dµ(x) = ρ(x)dm (x) with ρ(x) ∈ C(Rd), then Rsµ(x) exists for every d d x ∈ Rd. By C,c, possibly with indexes, we denote various constants which may depend only on d and s. We consider the following well-known conjecture. Conjecture 1.1. Let µ be a nonnegative finite Borel measure with compact support and continuous density with respect to the Lebesgue measure in Rd. There is a constant C such that sup |Rsµ(x)| ≤ C sup |Rsµ(x)|. (1.1) x∈Rd x∈suppµ For s = d−1 the proof is simple. Obviously, Rsµ(x) = ∇Us(x), (1.2) µ 1 2 VLADIMIREIDERMANAND FEDOR NAZAROV where 1 dµ(y) Us(x) = , s 6= 1, U1(x) = − log|y −x|dµ(y). µ s−1 Z |y −x|s−1 µ Z Thus each component of the vector function Rsµ(x), s = d−1, is harmonic in Rd \suppµ. Applying the maximum principle for harmonic functions we get (1.1). For d−1 < s < d, the relation (1.1) was established in [2] under stronger assumption that ρ ∈ C∞(Rd). In fact it was proved that (1.1) holds for each component of Rsµ with C = 1 as in the case s = d−1. The proof is based on the formula which recovers a density ρ from Us. But this method does not work for s < d−1. µ The problem under consideration has a very strong motivation and also is of independent interest. In [2] it is an important ingredient of the proof of the following theorem. By Hs we denote the s-dimensional Hausdorff measure. Theorem 1.2 ([2]). Let d−1 < s < d, and let µ be a positive finite Borel measure such that Hs(suppµ) < ∞. Then kRsµkL∞(md) = ∞ (equivalently, kRµsk = ∞). If s is integer, the conclusion of Theorem 1.2 is incorrect. For 0 < s < 1 Theorem 1.2 was proved by Prat [10] using different approach. The obstacle for extension of this result to all noninteger s between 1 and d−1 is the lack of the maximum principle. The same issue concerns the quantitative version of Theorem 1.2 obtained by Jaye, Nazarov, and Volberg [3]. The maximum principle is important for other problems on the connection between geo- metric properties of a measure and boundedness of the operator Rs on L2(µ) – see for µ example [3], [5], [6], [7]. All these results are established for d−1 < s < d or s = d−1. The problem of the lower estimate for kRsk in terms of the Wolff energy (a far going µ development of Theorem 1.2) which is considered in [3], [5], was known for 0 < s < 1. And the results in [6], [7] are (d −1)-dimensional analogs of classical facts known for s = 1 (in particular, [7] contains the proof of the analog of the famous Vitushkin conjecture in higher dimensions). For 0 < s ≤ 1, the proofs essentially use the Melnikov curvature techniques and do not require the maximum principle. But this tool is absent for s > 1. At the same time the validity of the maximum principle itself remained open even for 0 < s < 1. It is especially interesting because the analog of (1.1) does not hold for each component of Rsµ when 0 < s < d−1 unlike the case d−1 ≤ s < d – see Proposition 2.1 below. Weprove Conjecture 1.1 for0 < s < 1in Section2 (Theorem 2.3). The proofiscompletely different from the proof in the case d − 1 ≤ s < d. In Section 3 we prove Conjecture 1.1 in the special case of radial density of µ (that is when dµ = h(|x|)dm (x)), but for all d s ∈ (0,d). Section 4 contains an example showing that Conjecture 1.1 is incorrect for non- positivemeasures, evenforradialmeasureswithC∞-density(notethatin[14,Conjecture7.3] Conjecture 1.1 was formulated for all finite signed measures with compact support and C∞- density). 2. The case 0 < s < 1 We start with a statement showing that the maximum principle fails for every component of Rsµ if 0 < s < d−1. ON THE MAXIMUM PRINCIPLE FOR THE RIESZ TRANSFORM 3 Proposition 2.1. For any d ≥ 2, 0 < s < d−1, and any M > 0, there is a positive measure µ in Rd with C∞-density such that sup |Rsµ(x)| > M sup |Rsµ(x)|, (2.1) 1 1 x∈Rd x∈suppµ where Rsµ is the first component of Rsµ. 1 Proof. Let E = {(x ,...,x ) ∈ Rd : x = 0, x2 + ··· + x2 ≤ 1}, and let E , δ > 0, be 1 d 1 2 d δ a δ-neighborhood of E in Rd. Let µ = µ be a positive measure supported on E with δ δ µ(E ) = 1 and with C∞-density ρ(x) such that ρ(x) < 2/vol(E ) ≤ C /δ. Then δ δ d |Rsµ(x′)| > A , where x′ = (1,0,...,0), 0 < δ < 1/2. 1 d On the other hand, for x ∈ suppµ integration by parts yields 1 δ |Rsµ(x)| < dµ(y)+ dµ(y) 1 Z |y −x|s Z |y −x|s+1 |y−x|<δ |y−x|≥δ µ(B(x,δ)) δ µ(B(x,r)) ∞ µ(B(x,r)) = +s dr +δ(s+1) dr δs Z rs+1 Z rs+2 0 δ C δd Cs δ rd Cδ(s+1) 2 rd−1δ d < C + dr + dr +Cδ. δ δs δ Z rs+1 δ Z rs+2 0 δ Here by C we denote different constants depending only on d, and B(x,r) := {y ∈ Rd : |y −x| < r}. We have 2 2 rd−1 δln , s = d−2, δ dr = δ 1 Zδ rs+2 (2d−s−2δ −δd−s−1), s 6= d−2. d−s−2 Thus, all terms in the right-hand side of the estimate for |Rsµ(x)| tend to 0 as δ → 0, and 1 we may choose δ and a corresponding measure µ satisfying (2.1). (cid:3) We need the following lemma. The notation A ≈ B means that cA < B < CB with constants c,C which may depend only on d and s. Lemma 2.2. Let µ be a non-negative measure in Rd with continuous density and compact support. Let 0 < s < d−1. Then for every ball B = B(x ,r), 0 ∞ dµ(B(x ,t)) (Rsµ·n)dσ ≈ rd−s−1µ(B)+rd 0 , (2.2) (cid:12)Z (cid:12) Z ts+1 (cid:12) ∂B (cid:12) r where n is the oute(cid:12)r normal vector(cid:12)to B and σ is the surface measure on ∂B. (cid:12) (cid:12) Proof. We will use the Ostrogradsky-Gauss Theorem and differentiation under the integral sign. To justify these operations and make an integrand sufficiently smooth, we approximate K(x) = x/|x|s+1 by the smooth kernel K in the following standard way. Let φ(t), t ≥ 0, be ε a C∞-function such that φ(t) = 0 as 0 ≤ t ≤ 1, φ(t) = 1 as t ≥ 2, and 0 ≤ φ′(t) ≤ 2, t > 0. Let φ (t) := φ(t), K (x) := φ (|x|)K(x), and Rsµ := K ∗µ. We have ε ε ε ε ε ε (Rsµ·n)dσ = ∇·Rsµ(x)dm (x) = e ∇·φ (|y −x|) y −x dµ(y) dm (x). Z∂B ε ZB ε d ZB(cid:20)ZRd ε |y−x|s+1 (cid:21) d e e The inner integral is equal to y −x y −x ∇·φ (|y −x|) dµ(y)+ ∇· dµ(y) =: I (x)+I (x). Z ε |y −x|s+1 Z |y −x|s+1 1 2 |y−x|≤2ε |y−x|>2ε 4 VLADIMIREIDERMANAND FEDOR NAZAROV One can easily see that ∂ y −x 1 1 1 C φ (|y −x|) < C + < , |y −x| ≤ 2ε. ε (cid:12)∂x (cid:20) |y −x|s+1(cid:21)(cid:12) (cid:20)ε|y −x|s |y −x|s+1(cid:21) |y−x|s+1 (cid:12) i (cid:12) (cid:12) (cid:12) Hence, (cid:12) (cid:12) 1 2ε 1 |I (x)| < C dµ(y) < C dµ(B(x,t)) 1 Z |y −x|s+1 Z ts+1 |y−x|≤2ε 0 µ(B(x,2ε)) 2ε µ(B(x,t)) ≈ + dt. (2ε)s+1 Z ts+2 0 Since µ has a continuous density with respect to m , we have µ(B(x,t)) < A td as t ≤ 2ε < d µ,B 1, x ∈ B. Taking into account that s < d−1, we obtain the relation I (x)dm (x) → 0 B 1 d as ε → 0. R To estimate the integral of I (x) we use the equality ∇· x = d−s−1. Thus, 2 |x|s+1 |x|s+1 dµ(y) I (x)dm (x) = C dm (x) (cid:12)Z 2 d (cid:12) Z (cid:20)Z |y −x|s+1(cid:21) d (cid:12) B (cid:12) B |y−x|>2ε (cid:12) (cid:12) (cid:12) (cid:12) dmd(x) = C dµ(y) (cid:18)Z (cid:20)Z |y−x|s+1(cid:21) B(x0,r+2ε) B∩{|y−x|>2ε} dm (x) d + dµ(y) =: C(J +J ). ZRd\B(x0,r+2ε)(cid:20)ZB |y−x|s+1(cid:21) (cid:19) 1 2 Obviously, r td−1dt ≈ rd−s−1, |y−x | ≤ r, dmd(x) Z ts+1 0 ≈ 0 ZB |y −x|s+1 rd , |y−x | > r. |y −x |s+1 0 0 In order to estimate J1 we note that for sufficiently small ε, dm (x) dm (x) d ≈ d ≈ rd−s−1, y ∈ B(x ,r +2ε). Z |y −x|s+1 Z |y −x|s+1 0 B∩{|y−x|>2ε} B Hence, J ≈ rd−s−1µ(B(x ,r+2ε)). Moreover, 1 0 rd ∞ dµ(B(x ,t)) J ≈ dµ(y) = rd 0 . 2 ZRd\B(x0,r+2ε) |y −x0|s+1 Zr+2ε ts+1 Passing to the limit as ε → 0, we get (2.2) (cid:3) Now we are ready to prove our main result. Theorem 2.3. Let µ be a non-negative measure in Rd with continuous density and compact support. Let 0 < s < 1. Then (1.1) holds with a constant C depending only on d and s. Proof. Let us sketch the idea of proof. Let a measure µ be such that µ(B(y,t)) ≤ Cts, y ∈ Rd, t > 0. For Lipschitz continuous compactly supported functions ϕ, ψ, define the form hRs(ψµ),ϕi by the equality µ 1 y −x hRs(ψµ),ϕi = (ψ(y)ϕ(x)−ψ(x)ϕ(y))dµ(y)dµ(x); µ 2 ZZRd×Rd |y −x|s+1 the double integral exists since |ψ(y)ϕ(x)−ψ(x)ϕ(y)| ≤ C |x−y|. If we assume in addition ψ,ϕ that ψdµ = 0, we may define hRs(ψµ),ϕi for any (not necessarily compactly supported) µ R ON THE MAXIMUM PRINCIPLE FOR THE RIESZ TRANSFORM 5 bounded Lipschitz continuous function ϕ on Rd; here we follow [4]. Let suppψ ∈ B(0,R). For |x| > 2R we have y −x y −x x ψ(y)dµ(y) = + ψ(y)dµ(y) (cid:12)ZRd |y −x|s+1 (cid:12) (cid:12)ZRd (cid:20)|y −x|s+1 |x|s+1(cid:21) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) C C (cid:12) (cid:12) (cid:12)≤ (cid:12) |yψ(y)|dµ(y)= ψ . (cid:12) |x|s+1 ZRd |x|s+1 Choose a Lipschitz continuous compactly supported function ξ which is identically 1 on B(0,2R). Then we may define the form hRs(ψµ),ϕi as µ y −x hRs(ψµ),ϕi = hRs(ψµ),ξϕi + ψ(y)dµ(y) (1−ξ(x))ϕ(x)dµ(x). µ µ ZRd(cid:20)ZRd |y −x|s+1 (cid:21) The repeated integral is well defined because dµ(x) ∞ µ(B(0,t)) ∞ 1 ≤ C dt ≤ C dt. Z |x|s+1 Z ts+2 Z t2 |x|>2R 2R 2R Assuming that Theorem 2.3 is incorrect and using the Cotlar inequality we establish the existence ofapositive measureν such thatν hasnopointmasses, theoperatorRs isbounded ν on L2(ν), and hRs(ψν),1i = 0 for every Lipschitz continuous function ψ with ψdµ = 0. ν It means that ν is a reflectionless measure, that is a measure without point mRasses with the following properties: Rs is bounded on L2(ν), and hRs(ψν),1i = 0 for every Lipschitz ν ν continuous compactly supported function ψ such that ψdµ = 0. But according to the recent result by Prat and Tolsa [11] such measures do nRot exist for 0 < s < 1. We remark that the proof of this result contains estimates of an analog of the Melnikov’s curvature of a measure. This is the obstacle to extent the result to s ≥ 1. We now turn to the details. Suppose that C satisfying (1.1) does not exists. Then for every n ≥ 1 there is a positive measure µ such that n 1 sup |Rsµ (x)| = 1, sup |Rsµ (x)| ≤ . n n n x∈Rd x∈suppµn Let µ(B(x,r)) θ (x,r) := , θ := supθ (x,r). µ rs µ µ x,r We prove that 0 < c < θ < C. (2.3) µn The estimate from above is a direct consequence of Lemma 2.2. Indeed, for any ball B(x,r) (2.2) implies the estimate c rd−1 ≥ (Rsµ ·n)dσ ≥ Crd−s−1µ (B), d n n (cid:12)Z (cid:12) (cid:12) ∂B (cid:12) (cid:12) (cid:12) which implies the desired inequa(cid:12)lity. (cid:12) The estimate from below follows immediately from a Cotlar-type inequality sup |Rsµ (x)| ≤ C sup |Rsµ (x)|+θ n n µn x∈Rd (cid:2)x∈suppµn (cid:3) (see [8, Theorem 7.1] for a more general result). 6 VLADIMIREIDERMANAND FEDOR NAZAROV Let B(x ,r ) be a ball such that θ (x ,r ) > c = c(s,d), and let ν (·) = r−sµ (x +r ·). n n µn n n n n n n n Then x−x Rsµ (x) = Rsν n , θ (y,t) = θ (r y +x ,r t). n n(cid:18) r (cid:19) νn µn n n n n In particular, ν (B(0,1)) = θ (x ,r ) > c. Choosing a weakly converging subsequence of n µn n n {ν }, we obtain a positive measure ν. If we prove that n (a) ν(B(y,t)) ≤ Cts, (b) hRsν,ψi = 0 for every Lipschitz continuous compactly supported function ψ with ν ψdν = 0, R (c) the operator Rs is bounded on L2(ν), ν then ν is reflectionless, and we come to contradiction with Theorem 1.1 in [11] mentioned above. Thus, the proof would be completed. The property (a) follows directly from (2.3). For weakly converging measures ν with n θ < C we may apply Lemma 8.4 in [4] which yields (b). To establish (c) we use the µn inequality Rs,∗µ(x) := sup|Rµs,ε1(x)| ≤ kRsµkL∞(md) +C, x ∈ Rd, C = C(s), ε>0 for any positive Borel measure µ such that µ(B(x,r)) ≤ rs, x ∈ Rd, r > 0, – see [12, Lemma 2] or [13, p. 47], [1, Lemma 5.1] for a more general setting. Thus, Rsν (x) := ε n Rs 1(x) ≤ C for every ε > 0. Hence, Rsν(x) ≤ C for ε > 0, x ∈ Rd, and the non- νn,ε ε homogeneous T1-theorem [9] implies the boundedness of Rs on L2(ν). (cid:3) ν 3. The case of radial density Lemma 2.2 allows us to prove the maximum principle for all s ∈ (0,d) in the special case of radial density. Proposition 3.1. Let dµ(x) = h(|x|)dm (x), where h(t) is a continuous function on [0,∞), d and let s ∈ (0,d−1). Then (1.1) holds with a constant C depending only on d and s. Weremindthatfors ∈ [d−1,d)Conjecture1.1isprovedin[2]foranycompactlysupported measure with C∞ density. Thus, for compactly supported radial measures with C∞ density (1.1) holds for all s ∈ (0,d). Proof. Because µ is radial, by (2.2) we have c rd−1|Rsµ(x)| = (Rsµ·n)dσ d (cid:12)Z (cid:12) (cid:12) ∂B(0,r) (cid:12) (cid:12) (cid:12) ∞ dµ(B(0,t)) ≈ (cid:12)rd−s−1µ(B(0,r))+r(cid:12)d , r = |x|. Z ts+1 r Thus, µ(B(0,r)) ∞ dµ(B(0,t)) |Rsµ(x)| ≈ +r . rs Z ts+1 r Fix w 6∈ suppµ, and let r = |w|. If µ(B(0,r)) ∞ dµ(B(0,t)) ≥ r , rs Z ts+1 r ON THE MAXIMUM PRINCIPLE FOR THE RIESZ TRANSFORM 7 then there is r ∈ (0,r) such that {y : |y| = r } ⊂ suppµ and µ(B(0,r)) = µ(B(0,r )). 1 1 1 Hence, µ(B(0,r)) µ(B(0,r )) |Rsµ(w)| ≈ < 1 ≤ C|Rsµ(x )|, |x | = r . rs rs 1 1 1 1 If µ(B(0,r)) ∞ dµ(B(0,t)) < r , rs Z ts+1 r then there is r > r such that {y : |y| = r } ⊂ suppµ and µ(B(0,r)) = µ(B(0,r )). Hence, 2 2 2 µ(B(0,r )) µ(B(0,r)) ∞ dµ(B(0,t)) 2 < < r , rs rs Z ts+1 2 r2 and we have ∞ dµ(B(0,t)) |Rsµ(w)| ≈ r ≤ C|Rsµ(x )|, |x | = r . Z ts+1 2 2 2 r2 (cid:3) 4. Counterexample Given ε > 0, we construct a signed measure ν = ν(ε) in R5 with the following properties: (a) ν is a radial signed measure with C∞-density; (b) suppν ∈ D := {1−ε ≤ |x| ≤ 1+ε}; ε (c) |R2ν(x)| < ε for x ∈ suppν; |R2ν(x)| > a > 0 for |x| = 2, where a is an absolute constant. Here R2ν means Rsν with s = 2. Let ∆2 := ∆◦∆, and let 2/3, |x| ≤ 1, u(x) = 1 1 − , |x| > 1. |x| 3|x|3 Note that ∆( 1 ) = 0 and ∆2( 1 ) = 0in R5 \{0}. Hence, ∆2u(x) = 0,|x| =6 1. Moreover, |x|3 |x| ∇u is continuous in R5 and ∇u(x) = 0, |x| = 1. For δ ∈ (0,ε), let ϕ (x) be a C∞-function in R5 such that ϕ > 0, suppϕ = {x ∈ δ δ δ R5 : |x| ≤ δ}, and ϕ (x)dm (x) = 1 (for example, a bell-like function on |x| ≤ δ}). Let δ 5 U := u ∗ ϕ . ThenR∆2U (x) = 0 as x 6∈ D . Also, ∆U (x) → 0 as |x| → ∞. Hence, the δ δ δ δ δ function ∆U can be represented in the form ∆U = c( 1 ∗∆2U ) (here and in the sequel by δ δ |x|3 δ c we denote various absolute constants). Set dν = ∆2U dm . Then suppν ∈ D , and (b) δ δ 5 δ δ is satisfied. Since ∆( 1 ) = c , we have ∆U = c∆( 1 ∗∆2U ), that is U = c( 1 ∗∆2U )+h, |x| |x|3 δ |x| δ δ |x| δ where h is a harmonic function in R5. Since both U and 1 ∗∆2U tend to 0 as x → ∞, we δ |x| δ have 1 U = c ∗∆2U . δ δ (cid:18)|x| (cid:19) Thus, y −x R2ν(x) = c dν (y) = c∇U (x). δ δ Z |y −x|3 Obviously, ∇U = ∇(u∗ϕ ) = (∇u)∗ϕ , and hence max |∇U (x)| → 0 as δ → 0. On the δ δ δ x∈Dδ δ other hand, for fixed x with |x| > 1 (say, |x| = 2) we have lim |∇u∗ϕ | = |∇u(x)| > 0. δ→0 δ Thus, (c) is satisfied if δ is chosen sufficiently small. 8 VLADIMIREIDERMANAND FEDOR NAZAROV Remark. It is well-known that the maximum principle (with a constant C) holds for potentials K(|x−y|)dµ(y) with non-negative kernels K(t) decreasing on (0,∞), and non- negative finRite Borel measures µ. Our arguments show that for non-positive measures the analog of (1.1) fails even for potentials with positive Riesz kernels. In fact we have proved that for every ε > 0, there exists a signed measure η = η(ε) in R5 with C∞-density and such that suppη ∈ D := {1−ε ≤ |x| ≤ 1+ε}, |u (x)| < ε for x ∈ suppη, but |u ((2,0,...,0))| > ε η η b > 0, where u (x) := dη(y), and b is an absolute constant. η |y−x| Indeed, for the firstRcomponent R2ν of R2ν we have 1 ∂ 1 ∂ ∂ R2ν = c U = c ∗ (∆2U ) = cu , where dη = (∆2U )dm . 1 ∂x δ (cid:18)|x| ∂x δ (cid:19) η ∂x δ 5 1 1 1 References [1] D. R. Adams, V. Ya. Eiderman, Singular operators with antisymmetric kernels, related capacities, and Wolff potentials, Internat. Math. Res. Notices 2012, first published online January 4, 2012, doi:10.1093/imrn/rnr258. [2] V. Eiderman, F. Nazarov, A. Volberg, The s-Riesz transform of an s-dimensional measure in R2 is unbounded for 1<s<2, J. Anal. Math. 122 (2014), 1–23; arXiv:1109.2260. [3] B.Jaye,F.Nazarov,A.Volberg,ThefractionalRiesztransformandanexponentialpotential,Algebra i Analiz 24 (2012), no. 6, 77–123; reprinted in St. Petersburg Math. J. 24 (2013), no. 6, 903–938; arXiv:1204.2135. [4] B. Jaye, F. Nazarov, Reflectionless measures for Calder´on-Zygmund operators I: general theory, to appear in J. Anal. Math., arXiv:1409.8556. [5] B. Jaye,F. Nazarov,M. C.Reguera,X. Tolsa,The Riesz transformofcodimensionsmaller thanone and the Wolff energy, arXiv:1602.02821. [6] F. Nazarov,X. Tolsa,A. Volberg, On the uniform rectifiability of AD regular measures with bounded Riesz transform operator: the case of codimension 1, Acta Math. 213 (2014), no. 2, 237–321; arXiv:1212.5229,88 p. [7] F. Nazarov, X. Tolsa, A. Volberg, The Riesz transform, rectifiability, and removability for Lipschitz harmonic functions, Publ. Mat. 58 (2014), no. 2, 517–532,arXiv:1212.5431,15 p. [8] F. Nazarov, S. Treil, and A. Volberg, Weak type estimates and Cotlar inequalities for Calder´on- Zygmund operators on nonhomogeneous spaces, Internat. Math. Res. Notices 1998, no. 9, 463–487. [9] F. Nazarov, S. Treil, and A. Volberg, The Tb-theorem on non-homogeneous spaces, Acta Math. 190 (2003), no. 2, 151–239. [10] L. Prat, Potential theory of signed Riesz kernels: capacity and Hausdorff measure, Internat. Math. Res. Notices, 2004, no. 19, 937–981. [11] L.Prat, X.Tolsa,Non-existenceofreflectionlessmeasuresfor the s-Riesz transformwhen 0<s<1, Ann. Acad. Sci. Fenn. Math. 40 (2015), no. 2, 957–968. [12] M. Vihtil¨a, The boundedness of Riesz s-transforms of measures in Rn, Proc. Amer. Math. Soc. 124 (1996), no. 12, 3797–3804. [13] A.Volberg,Calder´on-Zygmund capacities and operators on nonhomogeneous spaces.CBMSRegional Conference Series in Mathematics, 100. AMS, Providence, RI, 2003. [14] A. L. Volberg, V. Ya. Eiderman, Nonhomogeneous harmonic analysis: 16 years of development, Uspekhi Mat. Nauk 68 (2013), no. 6(414), 3–58(Russian); translationin Russian Math. Surveys 68 (2013), no. 6, 973–1026. Vladimir Eiderman, Department of Mathematics, Indiana University, Bloomington, IN E-mail address: [email protected] Fedor Nazarov, Department of Mathematics, Kent State University, Kent, OH E-mail address: [email protected]